Jacek Moscinski

Transmission Rates and Performance of a Network of Computers

Transmission rates for two communication schemes: ping/pong and ring transfer have been measured on multiprocessors and clusters of workstations. Performance of a network of computers as a tool for numerically intensive computing has been evaluated on the basis of SLALOM and EP-NAS parallel benchmarks.

Implementation of Parallel Lattice Gas Program on Workstations under PVM

In the paper we present lattice gas automata program for simulation of fluid flows. The parallel algorithm was elaborated using mainly domain decomposition and was implemented on clusters of workstations under PVM. Dynamic load balancing was applied in order to efficiently exploit computer power during normal load.

Tuning the Performance of Parallel Programs on NOWs Using Performance Analysis Tool

This paper presents the functionality and use of an off-line tool for performance analysis of parallel programs running on networks of workstations. The tool consists of a set of facilities used for instrumentation, monitoring, trace data processing and visualisation. It was applied to tune parallel programs implementing a lattice gas automata method and a travelling salesman problem solution with genetic algorithms. A study of how coupling or splitting messages could influence application’s performance in terms of communication overhead and load imbalance is presented.

Evaluation of Parallel Application's Behavior in Message Passing Environment

In the paper, we present the structure and functionality of a performance analysis environment for parallel applications running on networked workstations in message passing environment (PVM, MPI). Instrumentation of a parallel program can be carried out with a graphical tool. An execution trace is processed using a trace handling language. A series of metrics and visualizations help to conclude about an applications behavior. As an example we show the results of applying the environment to performance analysis of two typical parallel applications: lattice gas automata and finite elements method.

Load-Balancing Procedure for Distributed Short-Range Molecular Dynamics

In the paper we present a simple dynamic load-balancing procedure applied in short-range 12/6 Lennard-Jones molecular dynamics (MD) simulation on clusters of computers. For study fluid and mixtures in microcapillaries or growth of optical fibres, the computational box is arranged as a long cylinder.

Lattice Gas Automata and Molecular Dynamics on a Network of Computers

Parallel algorithms and programs for lattice gas automata and molecular dynamics for fluid flow simulation were developed and investigated using PVM, Express, Network Linda and p4 on a network of workstations and a CONVEX C3210.

Storing Large Volumes of Structured Scientific Data on Tertiary Storage

As the scientific data become larger and larger, a method is needed to store and retrieve it in time- and cost-efficient way. NCSA HDF is a data/programming model widely used for storing and sharing scientific data. We present an approach to store large volumes of structured scientific data in HDF format on tertiary storage.

Performance Analysis Environment for Parallel Applications on Networked Worstations

We present an approach to the development of a performance analysis environment for parallel applications running on networked workstations under message passing environments (PVM, MPI). Instrumentation of a parallel program is carried out with a graphical tool. Then an application is compiled and linked. An execution trace is processed with tools using a trace handling language. A series of metrics and visualizations help to conclude about application behavior. As an example we show the results of applying the environment to performance analysis of a molecular dynamics simulation parallel program.

Monitoring of Performance of PVM Applications on Virtual Network Computer

We present a toolkit of monitoring and performance analysis facilities together with sample monitoring sessions and results of performance analysis of three parallel programs running under PVM on networked workstations. A trace file collected by Tape/PVM is converted into SDDF metaformat and then it is visualized with Pablo based tool. Performance is analyzed also with a set of metrics.

Parallelisation of a Code for Animation of Multi-object System

In this paper we describe the realization of a computer system designed for performing 3-D animation of sets of macroscopic objects. Realistic movements observed on the computer screen are achieved by running complex simulation in the background. This simulation is based on fundamental physical rules (e.g. preservation of momentum and energy, Newtons laws of dynamics) which ensure that generated scenes are conformable to a common sense of a viewer.